How Razorpay built India's largest payments platform from scratch

Executive overview

Accepting digital payments in India in 2014 was harder than accepting cash — a problem no one was solving. Harshil Mathur co-founded Razorpay to fix that, navigating a year of regulatory approvals before processing a single transaction.

Regulated industries impose the same compliance burden on every entrant, turning the friction into a durable moat. The company grew 40x between 2017 and 2020 with sub-$200k monthly burn, made an early UPI bet that unlocked enterprise customers overnight, and is now rebuilding the entire platform around AI before a startup forces it to.

In B2B, the moat is trust — and trust is built by picking up the phone.

Finding the problem and pivoting to the right customer

• Payments in India were built for large corporations, not developers or small businesses.
• First GTM target was educational institutes — institutions that simply charged the extra 1% as a fee and didn't care about digital collections.
• Parallel conversations with co-working startups showed genuine demand; pivoted entirely to that segment.
• Customers consistently said the same thing: "Yes, we have this problem. Nobody's solving it."
• Customer-driven energy — not investors or regulations — sustained conviction through 12+ months of zero revenue.

Regulation as a moat, not just a burden

• YC batch completed without a single live transaction; final bank approval came one week before Demo Day.
• Every competitor faces identical licensing requirements — the year-long process is a structural filter, not a temporary disadvantage.
• Difficult for non-obvious reasons (regulation, not product complexity) signals that customers will be loyal once you deliver.
• "If it's hard for other reasons, it's not really a problem — it's a moat."

The near-death moment: bank pulls the plug

• Two weeks after Demo Day, the partner bank shut down Razorpay's platform due to a single merchant complaint — 50+ live merchants went dark overnight.
• The team sat in a room and called every single affected merchant, explaining exactly what happened and what was being done.
• Many merchants abused them; the rule was to listen, never stop picking up the phone.
• A new banking partner was live within four to five days; several of those original merchants are still Razorpay customers today.
• The crisis produced a standing rule: if a support exchange exceeds three or four messages, call the customer — no exceptions.
• AI is now used across the business, but not in customer support: in fintech, support is a trust channel, not just a resolution channel.

Capital efficiency and the B2B logic

• Raised ~$11M at Series A with burn under $200k/month; put the remainder in fixed deposits earning more than the monthly burn.
• Investor complained the company was profitable — the market at the time expected high burn as a signal of ambition.
• B2B is inherently logical: a business pays for value added; remove the value and the customer leaves tomorrow.
• That intensity — having to justify value every day — is also what makes B2B defensible at scale: the relationship is earned continuously, not assumed.

The UPI bet that changed the company's trajectory

• UPI launched April 2016; two of India's largest banks hadn't integrated by the time demonetization hit in November 2016.
• Most payment gateways did not integrate UPI, reasoning that without the large banks it wouldn't scale.
• Razorpay went live on UPI in September or October 2016 — the first payment gateway in the country to do so.
• Demonetization triggered massive demand for UPI; Zomato, Swiggy, BookMyShow, and other large platforms migrated to Razorpay within weeks because no other gateway was ready.
• It took six months for any competitor to go live on UPI; that window allowed Razorpay to enter enterprise segments it could not have otherwise reached.
• Small players can make early bets because the downside is bounded and the upside is asymmetric.

Facing acquisition offers and staying independent

• Acquisition approaches came even before the public launch — major global payment companies noticed the YC announcement.
• The case for independence: global players underestimated India's complexity and were unlikely to invest at the depth required for a market growing from $60B to $180B+ (Razorpay's own processing volume today).
• Rule of thumb used: only sell if the outcome can be achieved faster with someone else — that condition was never met.

Founder mode vs. manager mode

• As the company scaled, Mathur shifted to managing leaders rather than staying close to the product — the classic "manager mode" trap.
• Leaders, however good, will never care about the company as much as the founder; product vision in particular cannot be delegated.
• The correction took nearly 10 years to fully recognise and act on.
• The principle: delegate everything except the decisions that require the founder's conviction.

Rebuilding for AI — acting before a startup forces the issue

• Leadership started by using AI tools personally (Claude Code, Cursor) to rebuild intuition for what the technology enables.
• Sat down and asked: "If we were starting Razorpay today, how would we build it?" — mapped onboarding, integration, support, and platform interactions from scratch.
• Relaunched the platform with AI at its core; early traction is strong, though the technology is still maturing.
• The incumbent fallacy: waiting to respond to market change means the market has already moved past you.
• AI compresses the build — the only remaining moat is knowing what to build and moving first.

Advice for founders

• AI makes it easier to ship, which makes it easier to latch onto problems you don't actually care about.
• The filter that matters: can you spend the next 10 years on this specific problem?
• Company building will still take 10 years regardless of how fast the tools get.
• Connect deeply with the problem first; the technology is a multiplier, not a substitute for conviction.